Respiratory data, simple clinical factors identify patients with non-small cell lung cancer

Important points:

• By including breath data in predictive models, researchers may be able to distinguish between people with lung cancer and those without.
• Knowledge of sex, age and pack years improved lung cancer identification.

A predictive model incorporating both electronic nose breath data and basic demographics was able to identify lung cancer in a cohort of patients, according to the results of the study, published in . chest.

“By combining breath data and clinical parameters in a multicenter, multidevice validation study, we are able to adequately distinguish between lung cancer patients and lung cancer-free subjects in a non-invasive manner.” sharina card, medical doctor, PhD, Professors and colleagues at Medisch Spectrum Twente Respiratory Medicine in Enschede, the Netherlands, write: “This study paves the way for the introduction of breath analysis into the routine practice of lung cancer diagnosis.”

In a multicenter validation study, Kort et al. evaluated the ability of predictive models using breath patterns and clinical factors to distinguish between lung and non-lung cancers in a cohort of 575 individuals with the following conditions: bottom. non-small cell lung cancernegative for suspected lung cancer, or healthy.

The researchers obtained respiratory data from all patients using an electronic nasal device, aeoNose (The eNose Company), measuring 5-minute tidal breaths and assessing volatile organic compounds (VOCs).

As Helio previously reported“VOCs are breakdown products of biochemical processes in the body, and changes in VOC composition may indicate medical conditions such as cancer.”

Of the entire cohort, 376 (160 with lung cancer, 51 with suspected lung cancer, 165 healthy) belonged to the training set where the model was established, and 199 (79 with lung cancer, 32 with suspected lung cancer). ) belonged to , lung cancer-negative, 88 healthy) belonged to the validation set. Of note, the validation cohort was blinded.

To test the diagnostic ability of the model, researchers evaluated the area under the receiver operating characteristic curve (AUC-ROC).

result

Using respiratory data only and a cutoff value of -0.36 for determining lung cancer, the researchers found that the model tested on the training set had an AUC-ROC of 0.83 (95% CI, 0.79-0.87), a sensitivity of 91%, and a specificity of 0.83. found to be 54%. Negative predictive value is 89%. A slightly lower value was observed in the validation cohort with an AUC-ROC of 0.79 (95% CI, 0.72-0.85), a sensitivity of 88%, a specificity of 52% and a negative predictive value of 87%.

Compared to models containing respiratory data only, models considering sex, age and number of pack years alone showed poor predictive ability in both the training cohort (AUC-ROC, 0.67; 95% CI, 0.61-0.72) and the validation cohort. was shown to be low. (AUC-ROC, 0.75; 95% CI, 0.68-0.82).

However, the researchers found that using a trained model that considered both respiration and simple clinical data in a multivariate logistic regression analysis improved diagnostic ability (AUC-ROC, 0.87; 95% CI , 0.83–0.9). Of note, the cutoff probability for lung cancer was 16%. Evaluating the sensitivity (95%), specificity (51%), and negative predictive value (94%) of the joint data model also showed improved or similar values ​​compared to the model containing only respiratory data. value was observed.

According to the researchers, the model validated with 0.86 (95% CI, 0.81-0.91) AUC-ROC, 95% sensitivity, 49% specificity, and 94% negative predictive value using the same cutoff probability. showed comparable performance on the set.

Finally, the Mayo Clinic’s nodule calculator tool takes into account “age, smoking history, previous malignancies, nodule size, upper lobe spinous processes, and PET scan results” to provide respiratory data, gender, age, Tested against models of research involving pack-ears. .

Due to the limited data required for the Mayo Clinic tool, only 55 patients with suspected lung cancer were included. The cutoff probability for lung cancer was again set at 16%, the researchers said.

The researchers found that their model was more sensitive than the Mayo Clinic model (94.6% vs. 83.8%), less specific (22.2% vs. 61.1%), and had a lower positive predictive value (71.4% vs. 81.6%) and negatives as well. Hit rate (66.7% vs 64.7%).

“Validation of the predictive model performed in this study is a crucial step for the clinical integration of breath analysis in the diagnostic process of lung cancer,” write Kort et al. “We envision using the aeoNose in parallel with current practice in hospital settings to assess the feasibility and acceptability of electronic noses in clinical settings.”

The researchers further wrote that despite being based on exploratory analyses, the validated model showed the ability to accurately discriminate between early-stage and non-lung cancers.

“This finding confirms that the model may have a role for screening purposes,” continued Kort et al. This should be investigated in the target population and perhaps a different cutoff value should be chosen. A wait-and-see strategy can also be used when there is suspicion (e.g., based on CT scans) and lung cancer is unlikely based on validated models. ”

The future of electronic noses

According to an accompanying editorial, this study by Kort et al. joins a growing body of literature on electronic noses as a potential tool for detecting lung cancer and improving the condition of patients with and without cancer. Dr. Michael N. Comer, Assistant Professor of Medical Research, Department of Allergy, Pulmonary, and Emergency Medicine, Vanderbilt University, and Dr. Rafael Paez, Clinical Fellow in the Department of Allergy, Pulmonary and Critical Care Medicine at Vanderbilt University.

“An eNose device with sufficiently high specificity could help rule out cancer in patients with incidentally detected pulmonary nodules who might undergo invasive diagnostic procedures,” said Kammer. Paez writes. “In this context, the eNose device could be the next step in management. Another future direction involves combining exhaled VOC analysis with other non-invasive biomarkers such as imaging and blood.” Finally, after external validation of previously locked and published VOC-based models, clinical utility studies will be needed to demonstrate improved patient-centered outcomes.”

reference: